FAULT 01Late detection
Peak demand windows are often identified only as they're happening.
By the time SCADA shows the overload, the cheapest options - storage pre-charge, contracted load shifts, scheduled imports - are already off the table.
AI energy load forecasting & grid optimization
See the peak before it hits the grid.
Prelko forecasts load continuously, flags peak-demand windows ahead of time, and recommends dispatch adjustments - so balancing the grid stops being reactive.
PEAK WINDOW
Peak forecastLead time 3 h
6:00-8:00 PM, 14% above baseline - pre-dispatch recommended. Simulated plan shaves 145 MW off the peak and keeps the feeder under capacity.
Approved by shift operator at 13:12
Metered load Forecast Confidence band Dispatch-adjusted plan Peak window Baseline
Illustrative network · 15-minute resolutionForecasts built from
SCADA & historian load data AMI and sub-meter reads Numerical weather feeds Solar & wind output Tariff and demand signals
The problem
Most control rooms meet the peak at the same moment the grid does.
Reactive balancing is expensive: spot purchases, emergency peaker starts, load shedding and penalty tariffs. The root cause is rarely capacity. It is timing - the peak was visible in the data, just not in time.
FAULT 02Manual lag
Manual dispatch planning can't keep pace with weather- and demand-driven load swings.
Day-ahead spreadsheets refresh once. Load moves every 15 minutes with temperature, humidity, holidays and industrial shift patterns.
FAULT 03Variability
Renewable variability makes forecasting harder, not optional.
Rooftop and utility solar hollow out the midday load, then leave a steep evening ramp. Forecasting gross load alone no longer describes the grid.
How Prelko works
One continuous loop, running every fifteen minutes.
Each new meter read, weather run or demand signal moves the loop forward. Nothing waits for tomorrow's planning meeting.
01
Forecast
Prelko models expected load continuously from historical patterns, weather data and demand signals.
02
Detect
Peak-demand windows and unusual load patterns are flagged ahead of time, with lead time and confidence attached.
Window18:00-20:00
vs baseline+14%
Lead time7 h
03
Recommend
Dispatch and balancing adjustments are proposed to smooth the anticipated peak, then routed to your operators.
Storage discharge-90 MW
Flexible load shift-55 MW
StatusOperator review
04
Report
Forecast accuracy and grid-performance trends are tracked over time, so the model and the team both get sharper.
What the AI actually does
Precise about its job. Explicit about its limits.
In scope
Forecasts short- and medium-term load
From 1 hour to 7 days ahead, at 15-minute resolution, using historical, weather and demand-signal data.
Flags peak windows and unusual deviations
Before they occur, with the expected magnitude, the lead time and how confident the model is.
Recommends dispatch and balancing adjustments
Including storage schedules and flexible load shifts that account for solar and wind variability.
Out of scope, by design
No direct control of grid infrastructure
Prelko has no write path to breakers, relays, inverters or plant controls.
No automatic dispatch execution
All recommendations route to your operators for approval and execution through your existing systems.
Recommend, then approveHow dispatch recommendations work
The model proposes. Your operator decides.
Every recommendation arrives as a structured, auditable card: which assets, which window, how many megawatts, and the expected effect on peak and capacity margin.
A
Issued with context
Forecast, peak window and the reasoning behind each action.
B
Approve, modify or decline
Operators adjust megawatts or timing before accepting.
C
Executed in your systems, logged in Prelko
Outcome is compared to forecast in the next report.
Awaiting approvalREC-0419 · issued 11:05 · expires 16:30
Pre-dispatch for forecast peak 18:00-20:00 on the Eastern feeder group
| Action | Window | MW |
|---|---|---|
| Charge battery storage, Substation 4 | 14:30-17:00 | +40 |
| Discharge battery storage, Substation 4 | 18:00-20:00 | -90 |
| Shift contracted flexible C&I load | 18:00-20:00 | -55 |
Peak before1,097 MW
Peak after951 MW
Capacity margin+129 MW
Approve Modify Decline Routed to shift operator
Built for both sides of the meter
Same forecasting engine. Two very different peaks.
A utility balances a whole network against capacity. A plant balances its own load against a demand charge. Prelko is configured for each.
Utilities & grid operators
Network-wide forecasts by substation and feeder, peak alerts ahead of capacity limits, and reporting that stands up to regulatory review.
Explore UtilitiesIndustrial & commercial operators
Facility and portfolio load forecasts that keep maximum demand under contract, schedule flexible processes, and line up with the utility's view of the grid.
Explore Industrial & CommercialNumbers that matter
Engineered around the interval your grid is measured in.
15min
Continuous forecasting
Load re-forecast at 15-minute resolution, updated as new data arrives.
168h
Rolling horizon
From the next hour out to seven days, so day-ahead and week-ahead plans share one model.
96/day
Intervals scored
Every interval is scored forecast versus actual, feeding ahead-of-time peak detection.
0auto actions
Operator in command
Renewable-aware recommendations, never executed without human approval.
Your load data stays yours, exportable at any time. Data & security
Forecasting engine
Five modules. One model of your load.
Start with forecasting on a single site. Add detection, dispatch, renewables and reporting as the network grows.
Load Forecasting
Rolling 15-minute forecasts, 1 hour to 7 days ahead, with confidence bands.
Learn morePeak Demand Detection
Peak windows and abnormal deviations flagged with lead time.
Learn moreDispatch Recommendations
Storage, flexible load and import adjustments for operator approval.
Learn moreRenewable Integration
Solar and wind variability folded into net-load forecasts.
Learn moreGrid Performance Reports
Accuracy, peak capture and balance trends over time.
Learn moreSee the full engine
How data flows from your meters and weather feeds to an approved dispatch.
Platform overviewForecast accuracy estimator
How far off is today's forecast? Size the gap.
Pick how you forecast now and the size of your network or facility. The panel returns an illustrative accuracy range in the same units your control room uses.
Current forecasting method
Actual Current method With Prelko
Illustrative accuracy gain
69%
lower forecast error (MAPE)
Peak-hour miss avoided
~51MW
less error at your peak interval
This is an illustrative estimate based on typical patterns, not a live analysis of your grid. Connect your load data for your actual numbers.